SCoRe: Submodular Combinatorial Representation Learning
This work addresses representation learning challenges, particularly in class-imbalanced scenarios, offering a versatile framework that unifies existing methods and provides incremental gains.
The paper tackles inter-class bias and intra-class variance in representation learning by introducing the SCoRe framework, which uses submodular information measures to formulate loss functions, resulting in improvements of up to 7.6% in classification on datasets like CIFAR-10-LT and 19.4% in object detection on IDD and LVIS.
In this paper we introduce the SCoRe (Submodular Combinatorial Representation Learning) framework, a novel approach in representation learning that addresses inter-class bias and intra-class variance. SCoRe provides a new combinatorial viewpoint to representation learning, by introducing a family of loss functions based on set-based submodular information measures. We develop two novel combinatorial formulations for loss functions, using the Total Information and Total Correlation, that naturally minimize intra-class variance and inter-class bias. Several commonly used metric/contrastive learning loss functions like supervised contrastive loss, orthogonal projection loss, and N-pairs loss, are all instances of SCoRe, thereby underlining the versatility and applicability of SCoRe in a broad spectrum of learning scenarios. Novel objectives in SCoRe naturally model class-imbalance with up to 7.6\% improvement in classification on CIFAR-10-LT, CIFAR-100-LT, MedMNIST, 2.1% on ImageNet-LT, and 19.4% in object detection on IDD and LVIS (v1.0), demonstrating its effectiveness over existing approaches.